Seminário especial - Metodológicas CEM

No dia 30 de setembro de 2016 (sexta-feira), das 9h30 às 11h30, acontece um seminário especial dentro da série Metodológicas CEM:  Simulação Social Baseada em Evidências (Evidence-based social simulation modelling). A conferência será proferida por Scott Moss (Manchester University) na Sala 8 do Prédio da Filosofia e das Ciências Sociais (FFLCH/USP), Av. Prof. Luciano Gualberto, 315. Cidade Universitária, Butantã. O evento é aberto a todos sem necessidade de inscrição prévia, sendo que a palestra será proferia em inglês.

Scott Moss é professor de Simulação Social e Diretor do Centro de Modelagem de Políticas Públicas (Centre for Policy Modelling) na Manchester Metropolitan University. É também fundador da European Social Simulation Association e publicou em períodos como American Journal of Sociology (AJS) e Journal of Artificial Societies and Social Simulation (JASSS).

A palestra está voltada especialmente para pesquisadores e estudantes de ciências sociais (Ciência Política e Sociologia), pós-graduandos em sistemas complexos, políticas públicas, meio ambiente e território, economia, bem como nas áreas de Ciências da Computação, Matemática e Estatística.

Resumo

For the past 30 years or so, the lecturer has been committed to declarative rule-based modelling of human behaviour.  The rules are expressed in natural-language-like terms. (If it is raining, I will carry an umbrella.  He gives an example: If I am attacked by a robber and I have my umbrella, I will hit the robber with my umbrella.)  Most agent-based social simulation modellers use numbers to represent intensity of feeling, conditions and actions.  Generally, numerical models are procedural: they are based on algorithms which determine the sequences of steps taken by the agents.  In declarative rule-based models, the sequence of conditions facing agents taken by them emerge during the course of the simulations. He will show why declarative, rule-based models are better and more usefully based explicitly on available evidence.  Generally, such models are more complicated than number-based models and they require more processing power and memory.  They run (usually much) more slowly.

The core issue is this:  procedural models are simple and fast and can incorporate large numbers of agents.  They are also relatively easy to understand.  Evidence-based declarative models are more difficult to understand because they are much more complicated (in addition to being slower).  However, they contain and produce much more information in a form that can be explored and understood by policy-makers and others who are not technically minded and whose expertise is in the domain of application rather than modelling and formal analysis.  It would be premature and probably wrong to argue that on of these modelling approaches is to be preferred over the other.  At least for now, we should be forming an understanding of the conditions in which one or the other of these approaches is most appropriate and whether and how they might complement one another.